A NEW PERSPECTIVE ON PROVIDING
CLOUD COMPUTING SECURITY
A Position Paper
Ashraf Matrawy
, Clifford Liem, Michael Wiener, Yuan Xiang Gu and Andrew Wajs
Cloakware Advanced Research Center (CARC), Irdeto, Ottawa, Canada
Keywords:
IPR, Cloud, Isolation, Multi-tenancy, Hypervisor, Mutual auditing, Extended trust, Perimeter defence.
Abstract:
Security issues in cloud computing are ranked high on the list of reasons why many organizations delay
considering the computing paradigm for their future plans. There have been numerous discussions about the
security problem in cloud computing. In this paper, we take a different stand and discuss the impact that
cloud computing will have on offering security. The paper points out the unique aspects of cloud computing
architecture and operation and cloud-specific security issues. These cloud-specific issues could have an impact
on a number of areas of offering information security. We discuss its impact on Intellectual Property Rights
(IPR) protection, organizations extending protection beyond the traditional notion of perimeter defence, and
on attack surfaces. We point out how providing security in the new computing paradigm may be affected and
that there might be some benefits that the cloud brings to the information security scene. In summary, in this
paper, we attempt to initiate discussions around these important issues.
1 INTRODUCTION
Cloud computing is not a new technology but rather
a new paradigm for delivering computing power to
different types of computing power consumers (Arm-
brust et al., 2010; Mell and Grance, 2009). While it
provides significant economic and technical benefits,
there are many security concerns surrounding it. This
new paradigm encompasses a multitude of computing
and communication technologies. Therefore, it is rea-
sonable to expect cloud computing to inherit most of
the security problems of these underlying technolo-
gies. Two cloud computing characteristic properties
contribute to the creation of security problems that are
distinct to this paradigm (1) IT outsourcing where an
organization’s data and software are hosted beyond its
control and (2) the sharing of physical resources by
multiple customers (competitors?) which is referred
to as multi-tenancy. There are other factors that may
create new problems but these two appear to be the
most prominent at the moment.
Many of the cloud computing security issues that
Ashraf Matrawy is an associate professor at Carleton
University. He did this work while working at CARC dur-
ing his sabbatical. After June 2011 he could be reached at
ashraf matrawy@carleton.ca.
have been reported recently (e.g. (ENISA, 2009;
CSA, 2009)) are not specific to cloud computing. Ac-
tually some reported cloud security problems date
back to the time-sharing era in the 1960s (Chen et al.,
2010). This paper focuses on the new aspects of cloud
computing that are creating security problems that are
specific to cloud environments. In particular, we focus
on how cloud computing will affect the offering of in-
formation security.
Contribution. We intend this work to be a starting
point for discussions about providing security in the
cloud rather than a paper listing security problems in
the cloud. This paper raises important questions about
crucial issues such as changing the notion of perime-
ter defence, how the cloud impacts Intellectual Prop-
erty Rights (IPR) protection, and the changing attack
surfaces. We also point out how the cloud may be
beneficial to security. Note that this paper addresses
the technical viewpoint only. Legal, financial, and
other aspects of the problem are outside the paper’s
scope.
Outline. This paper is structured as follows: Sec-
tion 2 briefly explains the special features of the cloud
environment that pertain to the discussion on the new
security problems in this paradigm. Section 3 briefly
650
Matrawy A., Liem C., Wiener M., Xiang Gu Y. and Wajs A..
A NEW PERSPECTIVE ON PROVIDING CLOUD COMPUTING SECURITY - A Position Paper.
DOI: 10.5220/0003449106500655
In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 650-655
ISBN: 978-989-8425-52-2
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
discusses the different efforts to summarize the cloud
security problems. Section 4 categorizes the new se-
curity problems that are special to the cloud. Section
5 discusses attack surfaces and Section 6 discusses
the changing notion of perimeter defence. Section 7
tackles IPR protection in the cloud while Section 8
focuses on security life cycle and the cloud. Section
9 concludes the paper.
2 THE CLOUD ENVIRONMENT
A comprehensive overview of cloud computing can
be found in the work by the UC at Berkeley (Arm-
brust et al., 2010). Examples of other overviews are
the presentation by NIST at ACM CCSW 2009 (Mell
and Grance, 2009) and the white paper by Sun Mi-
crosystems (Sun, 2009). In this section, a summary
of the special features discussed in these references is
presented.
The elastic nature of the cloud is a distinct char-
acteristic. The ability to scale horizontally is more
important than efficiency on a single Virtual Machine
(VM). This is especially important in cases where
“surge computing” (Sun, 2009) is needed. Applica-
tions could be programmed to request more virtual
resources from the elastic infrastructure. This could
be done through APIs specified by the cloud provider
(Sun, 2009). This means that creating and managing
VMs is crucial in cloud computing. The security of
this process is one of the important new aspects.
In terms of cost and exposure, clouds could be
private, shared within a certain community, or pub-
lic. This order reflects an increasing level of expo-
sure and a decreasing cost. Clouds could also be a
hybrid of the models above. In organizations requir-
ing a high degree of isolation (e.g., government), the
cost to have a private cloud might be justifiable. A
community cloud might be shared among organiza-
tions with similar requirements. We note that most of
the discussion in this paper applies to shared clouds
(public or community) and not the case where an or-
ganization runs its own private cloud.
Clouds provide services at different levels usually
defined as SaaS (Software as a Service), PaaS (Plat-
form as a Service), and IaaS (Infrastructure as a Ser-
vice). The boundaries between these services may not
be clear and their definition may change depending on
the provider and the customer of the service.
3 TRADITIONAL SECURITY
PROBLEMS
This massive concentration of resources and data in
the cloud presents a more attractive target to attackers.
The vulnerabilities that exist in the underlying tech-
nologies naturally migrate to the cloud. The work by
the Cloud Security Alliance (CSA, 2009) and ENISA
(ENISA, 2009) includes long lists of these security
problems.
Vulnerabilities. According to the ENISA report
(ENISA, 2009) there are numerous vulnerabilities
that could affect the cloud environment. The report
lists 31 cloud vulnerabilities and 22 additional vul-
nerabilities that are not specific to the cloud. Some
of these vulnerabilities are not technical. The most
noted technical ones include hypervisor vulnerabili-
ties, user and application isolation, internal probing
and encryption-related vulnerabilities.
Benefits. There are arguments that cloud-based de-
fences can be more robust, scalable and cost-effective
and that these defences will be a market differentia-
tor between cloud providers. Additionally they could
provide (ENISA, 2009): (1) standardized interfaces
for managed security (2) rapid scaling of security
resources to activities such as filtering and encryp-
tion. Another activity is auditing and evidence gath-
ering through images of VMs although this aspect of
cloud security is proving to be very challenging (3)
timely and effective updates (4) resource centraliza-
tion (cheaper, easier, etc.).
However, some of the above mentioned benefits
could be the source of problems as well such as the
case of centralization where it becomes a single point
of failure.
Another benefit would be making some costly
security features more affordable where the cloud
provider makes up-front investment in a certain tool
that will make the cost per-user per-transaction much
more affordable for smaller users. An example is
forensic investigation tools (Molnar and Schechter,
2010).
4 CLOUD-SPECIFIC SECURITY
PROBLEMS
4.1 IT Outsourcing Problems
In the cloud, an organization will rely on applications
and systems that are not under its IT staffs control
and that may not fully comply with its own policies.
A NEW PERSPECTIVE ON PROVIDING CLOUD COMPUTING SECURITY - A Position Paper
651
Moreover, the organization’s data will be hosted out-
side. This has the potential to create security prob-
lems in the following areas:
Process. The systems used could be applying
risky approaches that could be unacceptable in
certain industries. Examples are the level of re-
quired testing, frequency of backups, etc.
Data. The loss of control of data creates obvious
security and privacy challenges.
Compliance/Legal Issues. Certain industries re-
quire more rigorous IT processes than others.
This could create new challenges in compliance
and perhaps complex legal issues.
Extended Trust. Users will have to trust
providers. They actually have to extend this trust
to multiple (chain of) providers. The issue of ex-
tended (chain of) trust (Chen et al., 2010) arises
from the fact that users will have to use an ap-
plication that is offered by a SaaS provider, who
in turn runs it on a PaaS provider running on a
VM provided as IaaS. The end user does not have
much of a choice but to trust the security in this
chain.
4.2 Multi-tenancy
Another main issue is the sharing of physical re-
sources where multi tenants (customers who could
even be competitors) are sharing the same physical
machine or storage system. The following are some
of the security issues that arise from multi-tenancy:
Data Deletion. Due to the shared data storage
resources, safe deletion problems are a concern.
VM Attacks.. Compromising one VM could lead
to the compromise of other VMs on the same
physical machine. An example of this is the work
on VM-to-host escape presented by Kortchinsky
(Kortchinsky, 2009) in Blackhat 2009.
Side/Covert Channel Attacks. Another inter-
esting attack arising from mutli-tenancy is pub-
lished by Ristenpart et al. (Ristenpart et al., 2009).
Side and covert channels could occur between two
VMs existing on the same physical server. As
demonstrated (Ristenpart et al., 2009) on EC2, it
is feasible to co-locate an attacking VM on a spe-
cific server that hosts a target VM. Attacks could
be mounted by measuring the cache load which
could be used to create side channels to verify
VM co-location on the same server, estimate traf-
fic rates, launch keystroke timing attacks, and in
general trace co-tenants’ activities. Chen et al.
(Chen et al., 2010) argue that tenant activity (not
only data) is something that requires protection as
it could reveal confidential information about the
users.
4.3 Auditing and Forensics
Mutual Auditing. Cloud computing security
problems are harder to investigate. Accountabil-
ity, third-party auditing, and reputation sharing is-
sues will continue to be a major obstacle to its de-
ployment. There is a need to create mutual audit-
ing techniques (Chen et al., 2010) which is unique
to this paradigm.
New Forensics Capabilities. Tools are needed
to trace problems and help with legal investiga-
tions. Problems could arise when one (or more)
system(s) is involved in a legal problem; all users
of the same system might be disrupted when this
system is under investigation.
5 ATTACK SURFACES
Cloud computing will result in the introduction of
new attack surfaces and will change other traditional
ones. An attempted taxonomy of attack surfaces is
presented by Gruschka and Jensen (Gruschka and
Jensen, 2010). The taxonomy defines six attack sur-
faces among three different entities: the user, the ser-
vice, and the cloud. We map the taxonomy to the three
entities
2
described in the Berkeley report (Armbrust
et al., 2010). Surfaces 1 and 2 are what normally hap-
pens in a client-server environment between the SaaS
user and the Software provided as a service. Sur-
face 3 (from the SaaS provider to the cloud infrastruc-
ture) represents all the attacks that an application can
mount against the infrastructure (hypervisor, VMs,
network). Surface 4 (a very critical one) includes the
attacks a cloud provider can perform against its direct
users (SaaS providers) which could be very harmful
given the control that the infrastructure provider has
over the VMs hosting the applications. Surfaces 5 and
6 represent indirect relationships between the SaaS
user and the cloud infrastructure provider since there
always exists a service provider (SaaS) in between the
two entities.
In non-cloud environments, attackers benefit from
the existence of a large number of unprotected home
machines that are running a full scale OS with many
unneeded services enabled. Will the cloud change
this by increasing the number of notebooks running a
2
Where user is the SaaS user, service is the SaaS
provider, and cloud is the cloud infrastructure provider.
CLOSER 2011 - International Conference on Cloud Computing and Services Science
652
lightweight OS with minimal cloud access tools (e.g.
a browser)? Could this reduce the total number of
available exploits and the number of machines run-
ning them? Will this reduce the success chances of
new worms? Will it be more difficult for botnets?
Will this force attackers to move to the heavily forti-
fied clouds where the providers invested appropriately
in security (hopefully)? It is still to be seen whether it
is going to easier or harder for attackers in these new
environments.
6 EXTENDING PROTECTION
BEYOND TRADITIONAL
PERIMETER DEFENCE
The notion of ”Perimeter Defence” has been a part
of information security that generated a significant
amount of discussion. An organization seeking to
protect its information infrastructure will create lay-
ers of defence starting with protecting its “perime-
ter” which traditionally denotes what physically sep-
arates its information infrastructure from the outside.
Tools for protecting the perimeter include firewalls,
filtering capabilities at routers, and the use of a DMZ
(demilitarized zone) among others. The effective-
ness of perimeters has been challenged as boundaries
around an organization information infrastructure has
become blurred. This in part is due to that fact that
many employees connect from outside and with tech-
nologies such as HTTP that could allow attackers to
pass through firewalls undetected.
In the cloud, an organization will have little con-
trol over how and where its data and applications are
hosted. The notion of perimeter defence as known to-
day will be challenged even more. In this section, we
ask the question: does moving an organization’s data
and applications into the cloud eliminate its capabil-
ity to keep its perimeter defence at all?
At this point, the answer seems to be that the no-
tion of perimeter defence will change dramatically in
the cloud. An organization could extend its security
protection to complement what the Cloud Provider
(CP) offers and not rely solely on the CP’s pro-
vided security. Even when the CP offers all kinds
of promised security, an organization that is serious
enough about its security may want to provide addi-
tional protection to what the CP provides to extend its
protection beyond its traditional perimeter. This could
mean extending protection
3
beyond its perimeter to
3
For example: to maintain data privacy and integrity,
protect applications from tampering, and even to conceal its
application activities from competitors.
include various parts of the cloud that will dynami-
cally change both in location and in nature (software,
platform, and infrastructure).
In all cases, it is expected that the CP will offer
some security tools and mitigation services. But how
can an organization extend its protection differently
depending on the type of cloud service provider?
The Case of IaaS. When asking the CP (IaaS
provider such as Amazon or GoGrid) for a vir-
tual machine, the CP is expected to secure the
network, the host OS, the hypervisor and isolate
VM instances. The CP could also provide the
clients with tools to use and customize. For exam-
ple, Amazon offers Multi-Factor Authentication,
customizable firewall, secure APIs, and Key man-
agement tools (Amazon, 2010). The organization
could extend its protection beyond its perimeter
to all the virtual machines that it has control over
in the cloud. Leveraging the organization’s secu-
rity tools and expertise could be done in a num-
ber of ways including securing the guest OS, en-
suring compliance with industry-specific require-
ments, creating multiple layers of security around
the VM, etc.
The Case of PaaS. If the required service from
the cloud is a platform where an organization de-
ploys its applications, the CP (PaaS provider) in
this case should protect the network and the plat-
form. The extended perimeter is this case protects
the application within this provided platform. An
example of such an environment is Google APP
Engine. The platform offers a protected sandbox
to the application with some restrictions on what
could be accessed. However, it offers customers
the option to integrate with a third-party Single
Sign-On provider (Google, 2011). Another ex-
ample is Microsoft Azure which offers identity
management, authentication, and other mitigation
services. However, for some threats, the appli-
cation has to implement its own mitigation (Mar-
shall et al., 2010).
Suppose that the application provider does not be-
lieve that the CP provides sufficient security. This
begins to raise the question of where more protec-
tion can possibly come from. The cloud offers the
ability to scale at limitless proportions; however,
all the application images are exactly the same.
Once an attacker succeeds with one deployed in-
stance, it is a simple extension to create an auto-
matic exploit to attack the rest.
The Case of SaaS. This is the situation where the
service sought from the cloud is solely an appli-
cation such as moving the enterprise email to a
A NEW PERSPECTIVE ON PROVIDING CLOUD COMPUTING SECURITY - A Position Paper
653
Cloud with Homogeneous Application Distribution
Cloud with Diverse Application Distribution
?
Protection on Perimeter?
Protection through Differing Instances
Attacker
Attacker
Figure 1: An Example of Protection by Diversity.
cloud-based email or managing human resources
using a cloud-based service. In this case, the
perimeter will extend to contain the data wher-
ever it is hosted. This scenario gives the customer
the lowest level of control. The importance of the
data that resides in the cloud could range from in-
cidental to competitor-critical to mission-critical.
The possible attacks could range from automated
searches to outright copying of data-base mate-
rial. For data protection, homomorphic encryp-
tion (Gentry, 2009) has been suggested but there
has been no practical solutions based on it yet.
Related work on practical techniques could be in
the area of white-box cryptography (Chow et al.,
2002; Wyseur, 2009).
At this point, what are the possibilities for extra
security beyond what the CP provides? Certainly,
data encryption can play a part for mission critical
information; however, is this overkill and potentially
slow for other types of data? Can we consider a con-
trasting approach where data populations are encoded
and varied, just enough to frustrate casual attacks and
categorically defeat the automated attack. When di-
versity is applied to an application population in the
cloud, it abates the threat that is accentuated by the
ability to scale. Also in terms of forensics, diver-
sity could help determine the location from which the
source code had been lifted from the cloud. Refer to
Fig. 1 for an illustration of diversity.
Any third party security provider (i.e., anyone
other than the CP, regardless of what type of CP) in
the cloud will have to deal with the problem of lack
of control and/or access to the cloud. Furthermore,
the possibility of rapid scaling and nearly limitless re-
sources emphasizes this lack of control. Moreover, if
a security problem arises, investigations will be lim-
ited to what the CP allows (Section 4.3). Given the
nature of these three different environments discussed
above, an IaaS environment provides its customers
with more control over PaaS in terms of investigation
capabilities. PaaS in turn provides more than SaaS.
7 PROTECTING INTELLECTUAL
PROPERTY RIGHTS
It is expected that a great amount of digital content
will eventually be placed in the cloud. There have
been ongoing discussions on different ownership, ju-
risdiction, financial, and protection issues related to
IPR in the cloud. It is expected the above issues will
cause confusion and situations will arise where own-
ers are not aware of where their content is hosted and
what laws govern its access and protection. This sec-
tion, however, focuses only on the technical issues
related to IPR protection in the cloud. Reference to
other issues could be found in (ENISA, 2009; CSA,
2009).
Content Exposure. The first question that could be
asked about IPR is about exposure. Is the cloud in-
creasing content exposure at the risk of increasing
abuse? The answer depends on the type of content
and the method of distribution. For a game that is
traditionally distributed on a media (e.g., a CD) that
the consumer buys and keeps, perhaps hosting this
game in the cloud will reduce the end user access to
the entire game. Reduced access to the entire game
may allow cloud-based manufacturers an advantage
to those that need to provide protection to a native ap-
plication on a desktop. On the other hand, for con-
tents that the owners traditionally offers from their
own servers, moving to the cloud might increase its
exposure. These servers will no longer be under the
owners’ control and this will increase chances of at-
tacks on content where the servers are moved to the
cloud.
Infringement Detection and Tracing. It is not
clear if the cloud will bring any advantage to the area
on IPR infringement detection and tracing. Due to the
fact that content owners will have little control over
the environment where content is hosted (or run in the
case of software), owners will have limited abilities
in detecting infringements. In other words, it might
not be better than the case where the content was dis-
tributed on media. Even with the help of the CP, it
may not be any easier since auditing and forensics ca-
pabilities for cloud environments are in a very early
stage. Infringement detection might be done manu-
ally and using human expertise for some time before
automated monitoring systems could be developed for
the cloud.
CLOSER 2011 - International Conference on Cloud Computing and Services Science
654
8 SECURITY LIFECYCLE
In general, effective software security means more
than an initial attack resistance. It also requires con-
tinual and active protection including: monitoring
for attacks, breach-response, and mitigation. In the
cloud, attention to the security lifecycle could mean
facing some new challenges compared with running
on a single computer. For example:
Security lifecycle requirements may dictate the
consideration of attacks that may be launched
from the cloud environment itself. What kind
of security lifetime strategy could prevent these
types of attacks? For example, can updatable soft-
ware take into account the shared nature of the un-
derlying infrastructure?
How is attack monitoring and detection of a com-
promise hindered by the limited control that appli-
cations owners are permitted by cloud providers?
Will a CP be able to provide the necessary sup-
port for special security hooks for monitoring and
detection?
Is there any forethought for regular notification of
the trustworthiness of applications running in the
cloud? If so, who should do these validations of
trust? The CP, the application itself, or both?
Consider the detection of a breach. The remedial
action might be limited since the problem may
only be solvable with the CP’s intervention. The
application owner may only be able to stop the
application or prevent access to certain users un-
til the issue is resolved. This raises an interesting
relationship issue between the application owner
and the CP on renewability and restoring the in-
tegrity of the application.
9 CONCLUSIONS
In this paper, we attempted to raise questions on how
cloud computing will affect the way security is of-
fered. We discussed the cloud characteristics that cre-
ate new challenges for security providers. Organiza-
tions should think of new ways to protect their assets
and interests when they are deployed in the cloud.
This protection may differ depending on the type of
cloud environment (IasS, PaaS, or SaaS). Similarly,
new ways of protecting IPR in the cloud are needed.
While the cloud may bring some usability benefits
where CPs (Cloud Providers) are expected to secure
the cloud infrastructure, it also raises new questions
where users and application providers have less con-
trol over how and where their data and programs are
deployed. The future is uncertain at best.
REFERENCES
Amazon (2010). Amazon web services: Overview of secu-
rity processes. Amazon White Paper.
Armbrust, M. et al. (2010). A view of cloud computing.
Communications of the ACM.
Chen, Y. et al. (2010). What is new about cloud computing
security? Technical Rep. UCB/EECS-2010-5, Univ. of
California at Berkeley.
Chow, S. et al. (2002). White-box cryptography and an aes
implementation. In 9th Annual Workshop on Selected
Areas in Cryptography.
CSA (2009). Security guidance for critical areas of focus in
cloud computing v2.1. Cloud Security Alliance.
ENISA (2009). Cloud computing: Benefits, risks and rec-
ommendations for information security. The Euro-
pean Network and Information Security Agency.
Gentry, C. (2009). Fully homomorphic encryption using
ideal lattices. In Proc. of the 41st annual ACM sym-
posium on Theory of computing.
Google (2011). Saml single sign-
on (sso) service for google apps.
http://code.google.com/googleapps/domain/sso/saml
reference implementation.html.
Gruschka, N. and Jensen, M. (2010). Attacksurfaces: A
taxonomy for attacks on cloud services. In IEEE 3rd
International Conf. on Cloud Computing.
Kortchinsky, K. (2009). Cloudburst, a vmware guest to host
escape story. A talk in BlackHat USA.
Marshall, A. et al. (2010). Security best practices for devel-
oping windows azure applications. Microsoft White
Paper.
Mell, P. and Grance, T. (2009). Effectively and securely
using the cloud computing paradigm. In ACM Cloud
Computing Security Workshop.
Molnar, D. and Schechter, S. (2010). Self hosting vs. cloud
hosting: Accounting for the security impact of hosting
in the cloud. In Workshop on Economics of Informa-
tion Security.
Ristenpart, T. et al. (2009). Hey, you, get off of my cloud:
exploring information leakage in third-party compute
clouds. In ’Proc. of ACM CCS.
Sun (2009). Introduction to cloud computing architecture.
Sun Microsystems White Paper.
Wyseur, B. (2009). White-box cryptography. In PhD the-
sis, Katholieke Universiteit Leuven, B. Preneel (pro-
motor), 169+32 pages.
A NEW PERSPECTIVE ON PROVIDING CLOUD COMPUTING SECURITY - A Position Paper
655